This slender and closely argued monograph examines the implications for democratic theory of the now well-known empirical and experimental evidence on framing effects on individual choices. Complete in itself, it is nevertheless rather reluctant to move out of its comfort zone.

The data on framing effects go back at least to Daniel Kahneman and Amos Tversky's work (actually they go back farther—the author could usefully have name-checked Maurice Allais and Daniel Ellsberg, for example). As most readers of Perspectives are aware, it is easy to get inconsistent, including cyclical, responses from experimental subjects by varying the frame in which a question is asked. Classically, when asked to choose between two treatments to mitigate a disease that would otherwise kill 600 people, if the choice is framed as between “saving 200 people for sure” and “a treatment with a 1/3 chance of saving 600 people and a 2/3 chance of saving nobody,” most subjects choose the former. If the choice is framed as between “400 people will die for sure” and “a treatment with 1/3 probability that nobody will die and 2/3 probability that 600 will die,” most subjects choose the latter. The choices presented are of course identical. Framing effects have been shown to affect the decisions of people we would rather they did not, such as doctors and air traffic controllers.

In survey research, framing effects are equally well known and lead responsible survey companies to be extremely careful with their question wording. The same effects give irresponsible politicians and interest groups a great deal of leeway to frame questions in order to get the answers they want. If anyone doubts this, he or she need only look at the huge range of reported opinion responses on such emotive questions as gun control and same-sex marriage.

How badly is democratic theory affected by these findings? And can a version of it be advanced that is robust to this reality? These are Jerome Terence Kelly's central themes. Kelly classifies normative democratic theory into multiple subsets and attempts to rank them in ascending order of their epistemic demands. According to his strategy here, the greater a theory's epistemic demands, the more it is damaged by the empirics of framing: “[T]he epistemic capacities of citizens are significantly less reliable than has been presumed by most normative theorists…. [We should] eliminate from our consideration theories of democracy that are unfeasible” (pp. 68, 72).

In Kelly's scheme, social-choice theory would be the least affected by humans' limited ability to think statistically. Kelly says at page 24 that “Arrow's formal result [is]aimed primarily, if not exclusively, at theories of democracy that interpret votes as voter preferences.” However, this is not what Kenneth Arrow thought he was doing, nor have most of his followers. Social choice is about a set of impossibility and existence results. It is for others, not for social-choice theorists, to decide what to do with them. Social-choice theory is not of itself a theory of democracy at all, and it should not, probably, have been in Kelly's list.

So what about epistemically demanding theories, such as all those derived from the Condorcet jury theorem? As readers will be aware, the jury theorem states that if the mean juror is (however slightly) more likely to make a correct than a wrong judgment on the facts, and if juror evaluations are mutually independent, then it is right to accept the majority judgment. There are obvious objections here (recapitulated by Kelly): Citizens' epistemic capacities are too feeble, real choices are not binary, jurors are not statistically independent, and so on. Are jury theorems dead in the water?

Not necessarily. Nate Silver got the 2012 presidential election result right; Karl Rove got it wrong. Why? Because Silver was a better user of the tools of statistical inference, including—most relevantly here—Bayesian updating. In some contexts, the wisdom of crowds works, especially in stock market situations in which there need only be a small number of well-informed people, good at Bayesian updating, to determine the market price of a good. It is now well established that political forecasts are among the goods that can be priced in this sort of “epistemic market,” if I may coin that phrase. So, in some contexts, even the most demanding epistemic theories can survive the cold blast of encountering the reality of framing effects and other pathologies of judgment.

Kelly's focus seems to be almost wholly on epistemic theories. So Framing Democracy has little to say about cases in which majority rule is just an aggregation of preferences, rather than of judgments, nor about cases in which majority rule applies to baseline value judgments (e.g., again, abortion or guns), where epistemic judgments can take you only so far. But within its self-imposed limitations, this remains a careful analysis of the theories it discusses.